MySQL 分割槽表 partition線上修改分割槽欄位

us_yunleiwang發表於2013-12-06

公司線上在用partition,有一個表的分割槽欄位錯了,需要重建,結果發現沒有辦法像修改主鍵欄位或者修改索引欄位那樣直接一條sql搞定。而是需要建臨時表,有down time,所以去仔細看了文件,研究下partition的細節問題。

自己公司線上採取的時候,凌晨1點業務低峰期,執行:

建立臨時表

CREATE TABLE tbname_TMP (
    SHARD_ID INT NOT NULL,
    ...

    xxx_DATE DATETIME NOT NULL,
    PRIMARY KEY (xxx_DATE,shard_id)
) ENGINE=INNODB DEFAULT CHARSET=utf8 COLLATE=utf8_bin
PARTITION BY LIST(MONTH(xxx_DATE)) (
    PARTITION m1 VALUES IN (1),
    PARTITION m2 VALUES IN (2),
    PARTITION m3 VALUES IN (3),
    PARTITION m4 VALUES IN (4),
    PARTITION m5 VALUES IN (5),
    PARTITION m6 VALUES IN (6),
    PARTITION m7 VALUES IN (7),
    PARTITION m8 VALUES IN (8),
    PARTITION m9 VALUES IN (9),
    PARTITION m10 VALUES IN (10),
    PARTITION m11 VALUES IN (11),
    PARTITION m12 VALUES IN (12)
);

切換表名字,修改表結構

RENAME TABLE xxx TO xxx_DELETED, xxx_TMP TO xxx;

匯入原始資料

insert into xxx select * from xxx_DELETEDxxx_DELETED;

OK,一切搞定,整個過程50分鐘,MMM failover切換中後outline操作表結構變更以及資料匯入,實際downtime不包括修改表結構分割槽欄位的時間,只包括failover切換時間 為30秒

MySQL Partition,看的官方英文資料,翻譯水平有限,有些不翻譯成中文了,直接貼英文了。
1 list partition table 
mysql> CREATE TABLE `eh` (
    ->   `id` int(11) NOT NULL,
    ->   `ENTITLEMENT_HIST_ID` bigint(20) NOT NULL,
    ->   `ENTITLEMENT_ID` bigint(20) NOT NULL,
    ->   `USER_ID` bigint(20) NOT NULL,
    ->   `DATE_CREATED` datetime NOT NULL,
    ->   `STATUS` smallint(6) NOT NULL,
    ->   `CREATED_BY` varchar(32) COLLATE utf8_bin DEFAULT NULL,
    ->   `MODIFIED_BY` varchar(32) COLLATE utf8_bin DEFAULT NULL,
    ->   `DATE_MODIFIED` datetime NOT NULL,
    ->   PRIMARY KEY (`DATE_MODIFIED`,`id`)
    -> ) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_bin
    -> /*!50100 PARTITION BY LIST (MONTH(DATE_MODIFIED))
    -> (PARTITION m1 VALUES IN (1) ENGINE = InnoDB,
    ->  PARTITION m2 VALUES IN (2) ENGINE = InnoDB,
    ->  PARTITION m3 VALUES IN (3) ENGINE = InnoDB,
    ->  PARTITION m4 VALUES IN (4) ENGINE = InnoDB,
    ->  PARTITION m5 VALUES IN (5) ENGINE = InnoDB,
    ->  PARTITION m6 VALUES IN (6) ENGINE = InnoDB,
    ->  PARTITION m7 VALUES IN (7) ENGINE = InnoDB,
    ->  PARTITION m8 VALUES IN (8) ENGINE = InnoDB,
    ->  PARTITION m9 VALUES IN (9) ENGINE = InnoDB,
    ->  PARTITION m10 VALUES IN (10) ENGINE = InnoDB,
    ->  PARTITION m11 VALUES IN (11) ENGINE = InnoDB,
    ->  PARTITION m12 VALUES IN (12) ENGINE = InnoDB) */;
Query OK, 0 rows affected (0.10 sec)


2 rang partition table 
mysql> CREATE TABLE rcx (
    ->     a INT,
    ->     b INT,
    ->     c CHAR(3),
    ->     d INT
    -> )
    -> PARTITION BY RANGE COLUMNS(a,d,c) (
    ->     PARTITION p0 VALUES LESS THAN (5,10,'ggg'),
    ->     PARTITION p1 VALUES LESS THAN (10,20,'mmmm'),
    ->     PARTITION p2 VALUES LESS THAN (15,30,'sss'),
    ->     PARTITION p3 VALUES LESS THAN (MAXVALUE,MAXVALUE,MAXVALUE)
    -> );
Query OK, 0 rows affected (0.15 sec)

 

3 create range use less character 
CREATE TABLE employees_by_lname (
    id INT NOT NULL,
    fname VARCHAR(30),
    lname VARCHAR(30),
    hired DATE NOT NULL DEFAULT '1970-01-01',
    separated DATE NOT NULL DEFAULT '9999-12-31',
    job_code INT NOT NULL,
    store_id INT NOT NULL
)
PARTITION BY RANGE COLUMNS (lname)  (
    PARTITION p0 VALUES LESS THAN ('g'),
    PARTITION p1 VALUES LESS THAN ('m'),
    PARTITION p2 VALUES LESS THAN ('t'),
    PARTITION p3 VALUES LESS THAN (MAXVALUE)
);

alter table structure,add a new partition block
ALTER TABLE employees_by_lname PARTITION BY RANGE COLUMNS (lname)  (
    PARTITION p0 VALUES LESS THAN ('g'),
    PARTITION p1 VALUES LESS THAN ('m'),
    PARTITION p2 VALUES LESS THAN ('t'),
 PARTITION p3 VALUES LESS THAN ('u'),
    PARTITION p4 VALUES LESS THAN (MAXVALUE)
);


4 List columns partitioning
character column
CREATE TABLE customers_1 (
    first_name VARCHAR(25),
    last_name VARCHAR(25),
    street_1 VARCHAR(30),
    street_2 VARCHAR(30),
    city VARCHAR(15),
    renewal DATE
)
PARTITION BY LIST COLUMNS(city) (
    PARTITION pRegion_1 VALUES IN('Oskarshamn', 'H?gsby', 'M?nster?s'),
    PARTITION pRegion_2 VALUES IN('Vimmerby', 'Hultsfred', 'V?stervik'),
    PARTITION pRegion_3 VALUES IN('N?ssj?', 'Eksj?', 'Vetlanda'),
    PARTITION pRegion_4 VALUES IN('Uppvidinge', 'Alvesta', 'V?xjo')
);

date column
CREATE TABLE customers_2 (
    first_name VARCHAR(25),
    last_name VARCHAR(25),
    street_1 VARCHAR(30),
    street_2 VARCHAR(30),
    city VARCHAR(15),
    renewal DATE
)
PARTITION BY LIST COLUMNS(renewal) (
    PARTITION pWeek_1 VALUES IN('2010-02-01', '2010-02-02', '2010-02-03',
        '2010-02-04', '2010-02-05', '2010-02-06', '2010-02-07'),
    PARTITION pWeek_2 VALUES IN('2010-02-08', '2010-02-09', '2010-02-10',
        '2010-02-11', '2010-02-12', '2010-02-13', '2010-02-14'),
    PARTITION pWeek_3 VALUES IN('2010-02-15', '2010-02-16', '2010-02-17',
        '2010-02-18', '2010-02-19', '2010-02-20', '2010-02-21'),
    PARTITION pWeek_4 VALUES IN('2010-02-22', '2010-02-23', '2010-02-24',
        '2010-02-25', '2010-02-26', '2010-02-27', '2010-02-28')
);

5 HASH Partitioning
 int column,it can use digital function
CREATE TABLE employeesint (
    id INT NOT NULL,
    fname VARCHAR(30),
    lname VARCHAR(30),
    hired DATE NOT NULL DEFAULT '1970-01-01',
    separated DATE NOT NULL DEFAULT '9999-12-31',
    job_code INT,
    store_id INT
)
PARTITION BY HASH(MOD(store_id,4))
PARTITIONS 4;

If you do not include a PARTITIONS clause, the number of partitions defaults to 1. as below:
CREATE TABLE employeestest (
    id INT NOT NULL,
    fname VARCHAR(30),
    lname VARCHAR(30),
    hired DATE NOT NULL DEFAULT '1970-01-01',
    separated DATE NOT NULL DEFAULT '9999-12-31',
    job_code INT,
    store_id INT
)
PARTITION BY HASH(store_id);

 date colum 
CREATE TABLE employees2 (
    id INT NOT NULL,
    fname VARCHAR(30),
    lname VARCHAR(30),
    hired DATE NOT NULL DEFAULT '1970-01-01',
    separated DATE NOT NULL DEFAULT '9999-12-31',
    job_code INT,
    store_id INT
)
PARTITION BY HASH( YEAR(hired) )
PARTITIONS 4;

truncate all data rows:  alter table rcx truncate PARTITION;

 

6 LINEAR HASH Partitioning
CREATE TABLE employees_linear (
    id INT NOT NULL,
    fname VARCHAR(30),
    lname VARCHAR(30),
    hired DATE NOT NULL DEFAULT '1970-01-01',
    separated DATE NOT NULL DEFAULT '9999-12-31',
    job_code INT,
    store_id INT
)
PARTITION BY LINEAR HASH( YEAR(hired) )
PARTITIONS 4;

Given an expression expr, the partition in which the record is stored when linear hashing is used is partition number N from among num partitions, where N is derived according to the following algorithm:
(1)  Find the next power of 2 greater than num. We call this value V; it can be calculated as:
     V = POWER(2, CEILING(LOG(2, num)))
     (Suppose that num is 13. Then LOG(2,13) is 3.7004397181411. CEILING(3.7004397181411) is 4, and V = POWER(2,4), which is 16.)
(2) Set N = F(column_list) & (V - 1). 
(3)  While N >= num: 
    Set V = CEIL(V / 2) 
    Set N = N & (V - 1)


 [註釋] & 在SQL裡面的計算原理為:比如
 把十進位制轉化進製成二進位制,就得到了 


 首先按右對齊,例如變成0011和1000,按照每一位的數字來判斷,如果兩個都是1,則結果的相應位置就是1,否則就是0
 如果是1011和1000,結果就是1000
 如果是0110和1010,則結果就是0010
 但是3是0011,8 是1000,所以3&8結果就是0

CEILING(X) CEIL(X): 返回不小於X 的最小整數值。
LOG(X) LOG(B,X) :若用一個引數呼叫,這個函式就會返回X 的自然對數。
POWER(X,Y) : 返回X 的Y乘方的結果值。


資料分佈在哪個片區的計算方法:
Suppose that the table t1, using linear hash partitioning and having 6 partitions, is created using this statement:

CREATE TABLE t1 (col1 INT, col2 CHAR(5), col3 DATE)
    PARTITION BY LINEAR HASH( YEAR(col3) )
    PARTITIONS 6;

Now assume that you want to insert two records into t1 having the col3 column values '2003-04-14' and '1998-10-19'. The partition number for the first of these is determined as follows:
 V = POWER(2, CEILING( LOG(2,6) )) = 8
 N = YEAR('2003-04-14') & (8 - 1)
    = 2003 & 7
    = 3
 (3 >= 6 is FALSE: record stored in partition #3)

The number of the partition where the second record is stored is calculated as shown here:
 V = 8
 N = YEAR('1998-10-19') & (8-1)
   = 1998 & 7
   = 6

 (6 >= 6 is TRUE: additional step required)

 N = 6 & CEILING(8 / 2)
   = 6 & 3
   = 2

 (2 >= 6 is FALSE: record stored in partition #2)


The advantage in partitioning by linear hash is that the adding, dropping, merging, and splitting of partitions is made much faster, which can be beneficial when dealing with tables containing extremely large amounts (terabytes) of data. The disadvantage is that data is less likely to be evenly distributed between partitions as compared with the distribution obtained using regular hash partitioning.

疑問之一:MySQL 如何用一條SQL,不需要用臨時表來刪除分割槽欄位?將分割槽表變成普通的表?

來自 “ ITPUB部落格 ” ,連結:http://blog.itpub.net/23490154/viewspace-1062506/,如需轉載,請註明出處,否則將追究法律責任。

相關文章